• It's not being trained right now. (https://www.theverge.com/2023/4/14/23683084/openai-gpt-5-rumors-training-sam-altman)
  • There's a tradeoff for using the compute for training vs inference. Inference for now has an immediate return; whereas training is a bet.
  • GPT-4 and GPT-3 were years apart.
  • We should first expect GPT-4.5.
  • I expect all the large companies trying to learn more during these new "era of LLM deployment" and learn lessons for future systems.
  • ...
Files
Files
Tip: Mention someone by typing @username